Abstract-We propose an improved channel estimator for reuse-1 orthogonal frequency division multiplexing (OFDM) cellular systems 1 . The proposed channel estimation technique exploits delay subspace structure in reducing the interference on channel estimation. The proposed pilot-based channel estimation technique initially estimates the multipath delay locations of both the desired and interference channels. In estimating multipath delays, we assume that the time-of-flight difference between the desired and interfering signals ensures that the multipath delay locations of the corresponding channels are distinct. This information is used to suppress interference in the multipathdelay domain, and define a channel interpolator with a lower normalized mean squared error (NMSE) when compared to the conventional modified least-squares technique (mLS). We also derive the analytical expression for the bit-error-rate of a zeroforcing (ZF) receiver based on the proposed channel estimator. In particular, we show that for uncoded OFDM, the match between the estimated BER and analytical BER is very good and the proposed estimator can outperform mLS by more than a order of magnitude in BER if the interference on the data subcarriers is significantly lower than the interference seen on the pilot subcarriers. Simulation results are also presented with turbo-coded OFDM which further demonstrates the efficacy of the proposed algorithm.